Robust real-time urban drainage control under communication failures: an information-aware multi-policy control...

Area of Science:

• Environmental engineering
• Water resource management
• Control systems engineering

Background:

• Urban drainage systems face increasing pressure from climate change and urbanization.
• Real-time control (RTC) mitigates combined sewer overflow (CSO) risks cost-effectively but is vulnerable to communication failures.
• Existing research primarily assesses passive robustness, neglecting active performance enhancement during communication disruptions.

Purpose of the Study:

• To propose and evaluate an information-aware multi-policy switching control framework for urban drainage systems.
• To enhance the performance and robustness of RTC under communication failures using neural networks.
• To address the limitations of single-policy control in dynamic urban environments.

Main Methods:

• Developed a multi-policy switching framework leveraging neural networks (NNs) for flexible input handling.
• Trained a set of NN-based control policies for different system state configurations.
• Selected appropriate policies dynamically based on information availability during simulated communication failures.
• Evaluated the framework on the Astlingen benchmark network with sensor disconnections and random communication failures.

Main Results:

• The proposed multi-policy switching framework consistently outperformed single-policy control in all simulated failure scenarios.
• Performance improvements increased with the duration of sensor disconnections.
• CSO volume was reduced by 3.7-9.6% under random communication failures compared to single-policy control.
• Gains were primarily attributed to better management of CSO onset and recession stages.

Conclusions:

• The information-aware multi-policy switching framework offers a robust and effective solution for managing urban drainage systems under communication failures.
• This approach enhances the exploitation of system regulation capabilities, particularly during critical phases of CSO events.
• The study highlights the potential of adaptive control strategies in improving the resilience of smart urban infrastructure.